AdaVLA: Adaptive Step Flow Matching for Training-free Acceleration of Vision-Language-Action Models
AdaVLA:用于无需训练的视觉-语言-动作模型加速的自适应步长流匹配
机构 * Sogang University(西江大学)
AI总结 本研究提出AdaVLA,一种无需训练的自适应框架,通过流匹配轨迹曲率度量动态调整推理步骤与MLP剪枝率,在LIBERO基准等测试中实现VLA模型加速且性能下降可忽略。
Comments Accepted to the 2026 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2026)